Abstract
According to the similarity-attraction theory, humans respond more positively to people who are similar in personality. This observation also holds true between humans and robots, as shown by recent studies that examined human-robot interactions. Thus, it would be conducive for robots to be able to capture the user personality and adjust the interactional patterns accordingly. The present study is intended to identify significant speech characteristics such as sound and lexical features between the two different personality groups (introverts vs. extroverts), so that a robot can distinguish a user's personality by observing specific speech characteristics. Twenty-four male participants took the Myers-Briggs Type Indicator (MBTI) test for personality screening. The speech data of those participants (identified as 12 introvertive males and 12 extroversive males through the MBTI test) were recorded while they were verbally responding to the eight Walk-in-the-Wood questions. After that, speech, sound, and lexical features were extracted. Averaged reaction time (1.200 s for introversive and 0.762 s for extroversive; p = 0.01) and total reaction time (9.39 s for introversive and 6.10 s for extroversive; p = 0.008) showed significant differences between the two groups. However, averaged pitch frequency, sound power, and lexical features did not show significant differences between the two groups. A binary logistic regression developed to classify two different personalities showed 70.8% of classification accuracy. Significant speech features between introversive and extroversive individuals have been identified, and a personality classification model has been developed. The identified features would be applicable for designing or programming a social robot to promote human-robot interaction by matching the robot's behaviors toward a user's personality estimated.